Statement of Interest
Total Page:16
File Type:pdf, Size:1020Kb
From: AAAI Technical Report WS-02-11. Compilation copyright © 2002, AAAI (www.aaai.org). All rights reserved. Statement of Interest: Prof James Hendler Director, Semantic Web and Agent Technologies Maryland Information and Network Dynamics Laboratory University of Maryland, USA http://www.cs.umd.edu/~hendler To understand this short article, I need you to protocol called the "HyperText Transfer use the following visualization technique. Protocol" (the "http" you have become so used Spread your hands about one foot wide. Imagine to). The creator of this program, Tim Berners- your right hand is the traditional AI knowledge Lee, realized that the hypertext systems representation view of formal ontologies. Think envisioned by his peers could never scale -- if of moving left as moving from carefully person A needed permission, or worse, had to constructed large ontologies to smaller pay, person B to link to a text, the system component ontologies, created in less formal wouldn't grow. Instead, he developed a scheme ways. About the time you hit your left hand that challenged the fundamental assumptions of you’re at the edge of the traditional AI view of the hypertext community, and his success has ontologies, and into something much less formal been legion! Further, the research communities and somewhat less well-defined. Now turn your doing markup and hypertext were profoundly head left and squint into the distance -- that's changed, and current meetings have few of the where my view of ontologies can be found!! In "old-timers" involved as major research this little piece let me see if I can explain this, participants. and help you realize that ontologies are going to What is the tidal wave threatening to swamp change the world, but only if we change our our community?TheSemantic Web effort, led view of ontologies! by a combination of researchers from It is clear to many of us [1] thatthefieldof government, industry and academia, wants to knowledge representation is about to get shaken take ontologies (and other KR technologies) and up and bent out of recognition in the way that the bring them totheweb.However, this means at a fields concerned with on-line documents were web scale! When these things start to proliferate, changed a few years back. The text markup and some of our basic notions of KR will be hypertext communities had spent years working challenged, and our community will lose if we on expressive text markup languages and try to kick in our heelsandstayfirm.For complicated programs designed to allow detailed example, on the web not only is there no way to analysis of online documents coupled together ensure consistency, but it is guaranteed that there by two-way links between the pages. Great care will be inconsistent information, information one was taken to guarantee that onewouldnever hit a doesn't believe, information that one cannot find link that "didn't work.” Guaranteeing consistency the provenance of, information you put there in the links required centralized repositories and being combined with information that others complex protocols to make sure if a document described (and which you may not approve of) was moved, the links movedwith it. The and all those other "web"things. Even worse, if community knew that the death of their systems I point at terms in your ontology, and then you would arise if links could be broken, and they change it (or move it somewhere else) my took great pains to avoid that! representationbecomes ungrounded -- KR has However, this assumption was blown out of never before had to deal with the AI equivalent the water by a small program called “World of Error 404! Wide Web” that was supported by a simple What will this brave new world of ontological information look like? Consider the that this is the web – different people will use the current web and the way it works. Documents of same semantic assertions for different uses, and many sizes and shapes are available all over the thus the exact form of a web language may not place, with some sites developed just to index matter as much as the fact that everyone uses it! others (some of the most popular sites on the I’ll even take that a step further – the web). These documents are linked in many semantic web will not be powered solely by different ways, and they refer to things on other “reasoning systems,” but by a wide variety of pageswith little or no effort required. Web tools and techniques that use ontologies as their ontologies will grow in that same way. People interlingua. These very different agents will who know that ontologies exist (the “web need to be able to trace terms found on different masters” of the semantic web) will help define pages and find common referents – the content ontologies for areas of interest to them. These will be distributed, inconsistent, and preferably ontologies will use terms from other ontologies (ready for this?) not veryexpressive!! On the and so on. In addition, products written for end- web expressivity is the kiss of death – simpler users, powered by access to these ontologies, languages and solutions go further than complex will also allow for the extension of ontological ones – HTML, a watered down version of the terms (for example, a form to be filled out might much more expressive SGML, took over the have a field called “other” which, when named, world. SGML developers are still wandering asserts a new property of the class being around telling people how everyone else is doing extended). it wrong, but strangely no one seems to listen. Now, with this in mind, let’s look at ontology And that, I believe, is the lesson we inthe languages. First and foremost, a web ontology KR community must take to heart. Web-based language must be anchored on the web. This ontologies are poised tomoveouttothe web and means that each “symbol” must be grounded at a become embedded in web tools (that is, most “universal resource indicator” (URI). Thus, on users will not even know they exist as they take the web there is no such thing as a generic advantage of the ontological content included in “person” filed, rather one must refer to a shrink-wrapped programts ranging from particular definition such as specialized process modeling tools like UMLS to http://www.cs.umd.edu/projects/plus/person- home page creation tools brought shrink- ontology.daml #person” which is in turn wrapped over the counter and business-related anchored to other web locations. This notion of applications like spread sheets and word web embedding is absolutely crucial to the processors. Note also that AI is already present semantic web, and thus it is a critical property. in many of these tools, but the usersaren’taware On top of the is the ability to relate these of it as they mouse the green lines to fix grounded terms one to another. Thus one can grammatical errors or use rule-based tax create classes and subclasses, use descriptions or preparation systems.) When this happens, KR subsumption, etc. However, it is crucial to keep researchers will have a choice – move with the in mind that this is the web, and that the web is times, or become the old ignored grouches huge! No reasoner can work by assuming it complaining about how the world simply doesn’t knows everything, has found all possible classes, understand what this KR stuff is all about. etc. without being able to assert what it is closed with respect to (a particular document or site, Tim Berners-Lee, Jim Hendler and Ora Lassila, the collection of facts found by a semantic web The Semantic Web, Scientific American, crawler, etc.) Further, it is worth remembering May, 2001..